Generative AI has triggered a wave of litigation and policy debates as artists challenge the unauthorized use of their work in model training. Major lawsuits, including those filed by artists Sarah Andersen, Kelly McKernan, and Karla Ortiz against Stability AI, Midjourney, and DeviantArt, center on allegations of copyright infringement and the opaque nature of training datasets. While some claims have faced dismissal in U.S. courts, the core legal battles regarding “dependency” and intellectual property rights remain active, forcing a global re-evaluation of how AI models must balance innovation with creator protections.
## The Legal Battle Over AI Training Data
The tension between generative AI developers and the creative community stems from the “black box” nature of machine learning. In January 2023, a group of artists initiated a class-action lawsuit against Stability AI, Midjourney, and DeviantArt. According to the court filings, the plaintiffs argue that these companies scraped billions of images from the internet without consent, effectively creating models that can mimic specific artistic styles without providing attribution or compensation to the original creators.
In the United Kingdom, Getty Images launched its own legal challenge against Stability AI. The stock photography firm alleged that its intellectual property was used without permission, citing instances where AI-generated images contained distorted versions of the Getty Images watermark. This case highlighted the broader issue of commercial AI platforms utilizing copyrighted material, a practice that has prompted international scrutiny from regulatory bodies.
## Judicial Hurdles and the Question of “Dependency”
Not all legal efforts against AI firms have succeeded in their initial stages. In October 2023, a U.S. federal judge dismissed parts of the class-action lawsuit filed by Andersen and her co-plaintiffs. The court ruled that the plaintiffs failed to provide sufficient evidence that the AI-generated outputs were “substantially similar” to the specific copyrighted works of the artists, a requirement for proving copyright infringement in this context.
Despite this setback, the litigation against Stability AI continues. A central point of contention in these proceedings is the concept of “dependency”—the extent to which an AI-generated work relies on a specific artist’s training data. Because most generative models do not track the provenance of individual training images, proving that a specific output is an unauthorized derivative remains a significant technical and legal hurdle.
## Industry Shifts Toward Commercial Safety
In response to these controversies, some companies have moved to create “commercially safe” AI models. Adobe introduced its Firefly model in 2023 with a focus on restricted training data. Unlike models that scrape the open web, Adobe stated that Firefly is trained exclusively on Adobe Stock images, public domain content, and works with expired copyrights.
Adobe’s approach includes a compensation model for contributors to Adobe Stock, providing a financial incentive for creators whose work is used to train the system. Furthermore, the company has integrated Content Credentials—based on the C2PA (Coalition for Content Provenance and Authenticity) standard—into its software. This metadata, which the company calls “digital nutrition labels,” aims to provide transparency by identifying whether a file was generated or edited by AI.
## Policy Responses and Regulatory Dialogue
The global debate has moved into government policy chambers, including Japan. In 2024, the Japanese Agency for Cultural Affairs convened the “Council for AI and Copyright,” a forum designed to bring together AI developers, rights holders, and creators. The objective is to define practical standards for how AI models should interact with copyrighted material.
Adobe has participated in these discussions, advocating for transparency as a cornerstone of AI ethics. By sharing its experience with Firefly’s training and the implementation of C2PA standards, the company has positioned itself as an intermediary between the rapid advancement of AI technology and the legal protections required by the creative sector. These policy dialogues represent a shift from purely reactive litigation to proactive rule-making, as nations attempt to establish a framework where AI innovation and human creativity can coexist.